An AI astrologer that reasons from a real chart
The model composes; the engine decides the facts.
Putting an LLM in front of astrology without a chart underneath it produces confident text that is not about anyone. The pattern that works is to compute the chart deterministically first and let the model write against those structured signals - so the answer is explainable, reproducible, and attributable to a planet rather than to a prompt.
6 endpoints across 3 stages.
Worth knowing first
Safety is a routing problem, not a prompt
Medical, financial, death-prediction and self-harm questions arrive in any consumer astrology surface. Handling them in the system prompt means one jailbreak away from disaster; these are intercepted before the model runs and answered with a grounded fallback.
Stage 1
One call for a grounded answer
Pass the birth details and the message. The chart, dashas and transits are computed first and the reply is composed against them - you do not assemble the context yourself.
Stage 2
Keep the thread
A caller-owned conversation id replays recent exchanges into the prompt, so follow-ups stay continuous without you tracking state.
Stage 3
Show your working
The chat surface is more convincing when the user can open the chart it is reasoning from. Same key, same envelope.
Start
One key covers all of it
Every endpoint above takes the same bearer token and returns the same envelope. Free tier covers 1,000 requests / month. Pick your language in the quickstarts, or run any of these without a key in the playground.
Other use cases